1. Field of the Invention
This invention relates to an image processing system which determines whether relevant pixels in a character image information obtained by reading an original on which characters have been recorded are object pixels to be subjected to enhancement processing (will be sometimes simply referred to as “object pixels”, hereinbelow) and carries out enhancement processing on pixels determined to be object pixels in sequence, thereby carrying out edge enhancement processing on the character image information.
2. Description of the Related Art
There has been proposed a technology of image processing in which character image information is obtained by reading an original, on which characters have been recorded, with a scanner or the like, whether relevant pixels in the character image information are object pixels to be subjected to enhancement processing is determined and enhancement processing is carried out on pixels determined to be object pixels, whereby edge enhancement processing is carried out on the character image information. By taking edge pixels (which are on the edge of a character) as the object pixels and carrying out enhancement processing on the edge pixels, a sharper character image can be output on the basis of the character image information.
Prior to carrying out the edge enhancement processing on the character image information, it is necessary to determine which pixel is an edge pixel. In Japanese Unexamined Patent Publication No. 10(1998)-240929, there is disclosed a method of determining which pixel is an edge pixel in which a difference in density between a relevant pixel and surrounding pixels is obtained and it is determined that the relevant pixel is an edge pixel when the difference is not smaller than a threshold value.
However the difference in density is smaller in character image information obtained by reading small point characters than in character image information obtained by reading large point characters as shown in FIG. 10. FIG. 10 schematically shows densities of pixels output when reading a part of a given character in the main scanning direction, for instance, with a scanner, the mountains showing the densities of parts corresponding to parts of the character and the valleys showing the densities of parts corresponding to parts of the background of the character. Accordingly, in order to distinguish the edge pixels on the edge of small point characters by the method disclosed in Japanese Unexamined Patent Publication No. 10(1998)-240929, it is necessary to set the threshold value smaller than the difference in density in character image information obtained by reading small point characters. However, since the difference in density in character image information obtained by reading small point characters is small and can be smaller than the difference between the density of the background of the edge portion of the original and the density of the outside of the original. In such a case, if the threshold value is set smaller than the difference in density in character image information obtained by reading small point characters, pixels on the border of the original can be also determined to be edge pixels and a line representing the edges of the original can be output.
In Japanese Patent No. 3115065, there is disclosed a method of determining which pixel is an edge pixel in which since pixels on the edge of a character in character image information obtained by reading small point characters generally have an intermediate density, the threshold value is set smaller when the relevant pixel is of an intermediate density and otherwise set at an ordinary value. However, this approach is disadvantageous in that pixels on the edge of the small point character cannot be distinguished from pixels on the border of the original since the intermediate density is defined without taking into account the density of the background of the edge portion of the original.
Though it is possible to enhance the edge pixels to such an extent that the contour of the original cannot be output, this approach is disadvantageous in that small point characters cannot be clearly output.